-
Notifications
You must be signed in to change notification settings - Fork 3
/
Train_precompiled_CNN.py
211 lines (165 loc) · 8.28 KB
/
Train_precompiled_CNN.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
# -*- coding: utf-8 -*-
"""
Created on Tue Feb 9 09:28:26 2021
@author: Patrice
"""
""" Libraries"""
import tensorflow as tf
from tensorflow.keras import regularizers
from tensorflow.keras import optimizers
from tensorflow.keras.callbacks import EarlyStopping
import numpy as np
import pandas as pd
import matplotlib.pyplot as plt
import pandas as pd
from tensorflow.keras.models import Sequential
from tensorflow.keras.layers import Dense, BatchNormalization, Flatten, Dropout
from tensorflow.keras.utils import to_categorical
from sklearn.model_selection import train_test_split
from sklearn import metrics
import sys
import os
from IPython import get_ipython #this can be removed if not using Spyder
'''User Inputs'''
TensorPath = '/media/patrice/DataDrive/SEE_ICE/JointTrain/JointTensor5k_T_float16.npy' #location of the tile tensor
LabelPath = '/media/patrice/DataDrive/SEE_ICE/JointTrain/JointTensor5k_L_float16.npy' #location of the labels
OutPath='/media/patrice/DataDrive/SEE_ICE/JointTrain/'
ModelTuning=False
TuningDataName='VGG16_50_Joint_RGBNIRfloat16_985acc'
NClasses=7
NIR=True
Tilesize=50
CNNname='VGG' # choose NAS for NASNET large, VGG for VGG16 or Dense for Densenet 121
batches=50 #try 200 for 50x50 tiles
targetacc=0.985
'''custom callback for accuracy target training'''
class MyThresholdCallback(tf.keras.callbacks.Callback):
def __init__(self, threshold):
super(MyThresholdCallback, self).__init__()
self.threshold = threshold
def on_epoch_end(self, epoch, logs=None):
accuracy = logs["val_accuracy"]
if accuracy >= self.threshold:
print('')
print('Validation accuracy target reached, stopping training')
print('')
self.model.stop_training = True
'''setup for RTX use of mixed precision'''
#Needs Tensorflow 2.4 and an RTX GPU
if ('RTX' in os.popen('nvidia-smi -L').read()) and ('2.4' in tf.__version__):
from tensorflow.keras import mixed_precision
mixed_precision.set_global_policy('mixed_float16')
from tensorflow.compat.v1 import ConfigProto
from tensorflow.compat.v1 import InteractiveSession
config = ConfigProto()
config.gpu_options.allow_growth = True
session = InteractiveSession(config=config)
'''Load Tensor Data'''
Tensor=np.load(TensorPath)
Label=np.load(LabelPath)
if NIR:
Tensor=Tensor[:,0:Tilesize, 0:Tilesize,:]
else:
Tensor=Tensor[:,0:Tilesize, 0:Tilesize,0:3]
'''Instantiate CNNs'''
if 'VGG' in CNNname:
# create cCNN model
inShape = Tensor.shape[1:]
from tensorflow.keras.applications import VGG16
CNN_base = VGG16(include_top=False,weights=None,input_shape=inShape)
model = Sequential()
model.add((CNN_base))
model.add(Flatten())
#Estimator.add(Dense(32,kernel_regularizer=regularizers.l2(0.001),kernel_initializer='normal',activation='relu'))
model.add(Dense(512, activation='relu',kernel_regularizer= regularizers.l2(0.001)))
model.add(Dropout(0.5))
model.add(Dense(256, activation='relu',kernel_regularizer= regularizers.l2(0.001)))
model.add(Dense(NClasses+1,kernel_initializer='normal',activation='softmax', dtype='float32'))
#Tune an optimiser
Optim = optimizers.Adam(lr=0.0001,beta_1=0.9,beta_2=0.999,decay=0.0,amsgrad=True)
# Compile model
model.compile(loss='categorical_crossentropy',optimizer=Optim,metrics=['accuracy'])
model.summary()
elif 'Dense'in CNNname:
from tensorflow.keras.applications import DenseNet121
inShape = Tensor.shape[1:]
CNN_base = DenseNet121(include_top=False,weights=None,input_shape=inShape)
model = Sequential()
model.add((CNN_base))
model.add(Flatten())
#Estimator.add(Dense(32,kernel_regularizer=regularizers.l2(0.001),kernel_initializer='normal',activation='relu'))
model.add(Dense(64, kernel_regularizer= regularizers.l2(0.001), kernel_initializer='normal', activation='relu'))
model.add(BatchNormalization(axis=-1, momentum=0.99, epsilon=0.001, center=True, scale=True, beta_initializer='zeros', gamma_initializer='ones', moving_mean_initializer='zeros', moving_variance_initializer='ones', beta_regularizer=None, gamma_regularizer=None, beta_constraint=None, gamma_constraint=None))
model.add(Dense(32, kernel_regularizer= regularizers.l2(0.001), kernel_initializer='normal', activation='relu'))
model.add(Dense(16, kernel_regularizer= regularizers.l2(0.001), kernel_initializer='normal', activation='relu'))
model.add(Dense(NClasses+1,kernel_initializer='normal',activation='softmax', dtype='float32'))
#Tune an optimiser
Optim = optimizers.Adam(lr=0.00001,beta_1=0.9,beta_2=0.999,decay=0.0,amsgrad=True)
# Compile model
model.compile(loss='categorical_crossentropy',optimizer=Optim,metrics=['accuracy'])
model.summary()
elif 'NAS'in CNNname:
from tensorflow.keras.applications import ResNet50
inShape = Tensor.shape[1:]
CNN_base = ResNet50(include_top=False,weights=None,input_shape=inShape)
model = Sequential()
model.add((CNN_base))
model.add(Flatten())
#Estimator.add(Dense(32,kernel_regularizer=regularizers.l2(0.001),kernel_initializer='normal',activation='relu'))
model.add(Dense(64, kernel_regularizer= regularizers.l2(0.001), kernel_initializer='normal', activation='relu'))
model.add(BatchNormalization(axis=-1, momentum=0.99, epsilon=0.001, center=True, scale=True, beta_initializer='zeros', gamma_initializer='ones', moving_mean_initializer='zeros', moving_variance_initializer='ones', beta_regularizer=None, gamma_regularizer=None, beta_constraint=None, gamma_constraint=None))
model.add(Dense(32, kernel_regularizer= regularizers.l2(0.001), kernel_initializer='normal', activation='relu'))
model.add(Dense(16, kernel_regularizer= regularizers.l2(0.001), kernel_initializer='normal', activation='relu'))
model.add(Dense(NClasses+1,kernel_initializer='normal',activation='softmax', dtype='float32'))
#Tune an optimiser
Optim = optimizers.Adam(lr=0.00001,beta_1=0.9,beta_2=0.999,decay=0.0,amsgrad=True)
# Compile model
model.compile(loss='categorical_crossentropy',optimizer=Optim,metrics=['accuracy'])
model.summary()
'''Tune and Fit CNN'''
Labels1Hot = to_categorical(Label)
def TuneModelEpochs(Tiles,Labels, model,TuningDataName,Path):
#Split the data for tuning. Use a double pass of train_test_split to shave off some data
(trainX, testX, trainY, testY) = train_test_split(Tiles, Labels, test_size=0.2)
history = model.fit(trainX, trainY, epochs = 50, batch_size = batches, validation_data = (testX, testY))
#Plot the test results
history_dict = history.history
loss_values = history_dict['loss']
val_loss_values = history_dict['val_loss']
epochs = range(1, len(loss_values) + 1)
get_ipython().run_line_magic('matplotlib', 'qt')
plt.figure(figsize = (12, 9.5))
plt.subplot(1,2,1)
plt.plot(epochs, loss_values, 'bo', label = 'Training loss')
plt.plot(epochs,val_loss_values, 'b', label = 'Validation loss')
plt.title('Training and Validation Loss for '+TuningDataName)
plt.xlabel('Epochs')
plt.ylabel('Loss')
plt.legend()
acc_values = history_dict['accuracy']
val_acc_values = history_dict['val_accuracy']
plt.subplot(1,2,2)
plt.plot(epochs, acc_values, 'go', label = 'Training acc')
plt.plot(epochs, val_acc_values, 'g', label = 'Validation acc')
plt.title('Training and Validation Accuracy')
plt.xlabel('Epochs')
plt.ylabel('Accuracy')
plt.legend()
plt.show()
DF=pd.DataFrame(history_dict)
DF.to_csv(Path+TuningDataName+'.csv')
sys.exit("Tuning Finished, adjust parameters and re-train the model")
if ModelTuning:
TuneModelEpochs(Tensor,Labels1Hot, model,TuningDataName,OutPath)
else:
(trainX, testX, trainY, testY) = train_test_split(Tensor, Labels1Hot, test_size=0.2)
when2stop = MyThresholdCallback(threshold=targetacc)
history = model.fit(trainX, trainY, epochs = 120, batch_size = batches, validation_data = (testX, testY), callbacks=when2stop)
Predictions=np.argmax(model.predict(testX), axis=1)
Observations=np.argmax(testY, axis=1)
report = metrics.classification_report(Observations, Predictions, digits = 3)
print('Classification report for '+TuningDataName)
print(report)
model.save(OutPath+TuningDataName+'.h5')
#close the TF session
session.close()